April 15th, 2009
|12:08 pm - Long-shot Prediction Game: the results|
A shamefully long time ago, I proposed a multi-stage prediction competition pertaining to twelve moderately unlikely events which might or might not happen, with a focus on the last five months of 2008 and the first month of 2009. Now we're well into 2009, the long-complete competition is starting to get rather rancid. Let me put it out of its misery before the stench of pallor completely stinks the place out.
First, we determine which - if any - of the twelve long-shots did come to fruition, like so. Again, I wll point to Iain's commentary on the matter. Returning to the original ordering, the twelve propositions were originally quoted as follows:
* Oil to trade at $200/barrel or higher.
This did not happen. I don't believe that any of Brent crude, New York light sweet crude or West Texas Intermediate crude traded at above $148/barrel. It's cute to note that this was a case where a long shot came to fruition, but the long shot was in the opposite tail of the curve to the one about which I enquired.
* BBC Global 30 stock market index to drop below 4750.
This did happen, dropping to lower than 4500 in October and as low as 4095.23 early last month. I don't claim this to be as reliable a barometer of the global economy as I might have liked it to be, but I'm not sure there are significantly better ones that could be tracked over the course of the year.
* John McCain to win at least 336 Electoral College seats in the 2008 US Presidential election.
This did not happen. There were no faithless electors and, as expected, John McCain won only 173 seats. (Well, votes; "seats" represents UK parliamentary vocabulary carried over in error.)
* Gordon Brown to stop being Prime Minister of the UK, Labour Party leader or both.
This did not happen. At times it seemed plausible, though.
* The USA to win at least one more gold medal than any other single country at the Beijing Olympics.
This did not happen. The USA won 36 gold medals, China won 51. (Now if I had specified medals of any colour, the USA would have outperformed all others, but I did not.)
* The New England Patriots to win Super Bowl XLIII, expected to take place on or around February 1, 2009.
This did not happen. The New England Patriots did not even make the play-offs, though as they had a better record than any other team not to make the play-offs, they might reasonably place themselves thirteenth at worst. Debates as to whether the 11-5 Patriots might be a more appropriate play-off team than, say, the 8-8 Chargers are possible.
* Microsoft to make a takeover bid for (or merger bid with) Yahoo that Yahoo's board of directors accept or recommend acceptance.
This did not happen. However, it's a story that still commands discussion and may be just a matter of time.
* Firefox usage to reach 33% or higher.
This did not happen. I declared three sources to be acceptable; the W3 counter reported 32.07% in September aggregating figures for versions 3.*, 2.* and 1.5 (so 1.5 to 126.96.36.199), but I don't think it would be reasonable to estimate usage of browsers from the 1.0 branch and from pre-1.0 branches to exceed 33% as required. TheCounter reported usage figures around 20% and Net Applications had figures around the same mark. Firefox usage increase seems to be decelerating, at least partially due to acceleration of usage of Safari and Chrome, and partially because MSIE 7 really isn't so bad any more.
* The Harry Potter and the Half-Blood Prince movie, i.e. Harry Potter 6, to gross at least $285,000,000 at the US box office by (and including) the weekend concluding Sunday February 1st, 2009.
This did not happen, because the release of the movie was postponed. Alas! Should the delayed release make it to US$285 million - plausible, because the upcoming July release might suit its brand of Fine Holiday Fun well - then a grumble would be entirely justifiable, though unsuccessful.
* One or more of Prince William, Prince Harry, Princess Beatrice and Princess Eugenie to marry publicly.
This did not happen. A new hat was not required.
* USA and/or Israel to execute an overt air strike against Iran.
This did not happen. Handy word, "overt". (I am not claiming that there were any covert ones, though...)
* There will be no more than two major hurricanes (category 3+) in the 2008 Atlantic Hurricane season.
This did not happen, sadly: Bertha was category three and Gustav, Ike, Omar and Paloma all were category four.
Thus, in conclusion, one (1) of the twelve long-shots happened. The prediction part of the game ran over two rounds: the first round saw people estimate the number of items which would happen over the time period, and the second round saw people effectively wager points on each of the twelve propositions at community-defined, and community-refined, odds. So now let's find out who won!
In the first round, players participated in two ways: firstly by estimating the number of events that would happen and secondly by suggesting which would be the most likely and least likely to occur.
The number of events that actually happened was one. Of the twenty entries to the first round, nineteen estimated the number of long-shots that would take place, with the estimates ranging from two to eight. "Well done, everybody!", as Chris Tarrant says, regardless of how well or badly the just-asked audience have actually done. The two players who nominated two events score nine points, the seven players who nominated three events score eight points and so on. Three bonus points are awarded to each of the nine players saying that the stock market drop was among the four most likely to happen; two penalty points are deducted from the single player saying that the stock market drop was among the four least likely to happen.
These bonus points make the difference, as of the two players who said that two long-shots would happen, one would nominate the stock market drop as among the most likely, scoring +3 points, and the other would nominate the stock market drop as among the least likely, scoring -2 points and eliminating herself from contention. Sorry, huskyteer; with a score of 12 points, the winner of the first round is Iain, who once wrote at daweaver. Well done, Iain!
In joint second place, with eleven points (eight for a nomination of three events plus three for proposing the stock market drop) were bateleur, quintus_marcius and jaydlewis. In joint fifth place, on ten points (seven for nominating four events plus three for the stock drop) were enfarcer, sir_quirky_k and malachan. Scores were as low as five, four and technically zero.
In the second round, players chose for each of the twelve propositions whether they would prefer to earn a large score in the unlikely case of the long-shot occurring or a small score in the likely case of the long-shot not occurring. Accordingly, the twelve propositions are scored like so:
I choose to earn 53 points if "USA to win most gold medals in Beijing" happens, instead of earning 47 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 47 points.
I choose to earn 55 points if "HP 6 to gross at least $285 million at the US box office" happens, instead of earning 45 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 45 points.
I choose to earn 57 points if "BBC Global 30 stock market index to drop below 4750" happens, instead of earning 43 points if it doesn't: players choosing to support the long-shot score 57 points, players choosing to oppose the long-shot score zero.
I choose to earn 59 points if "Oil to trade at $200/barrel or higher" happens, instead of earning 41 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 41 points.
I choose to earn 65 points if "Microsoft to takeover Yahoo" happens, instead of earning 35 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 35 points.
I choose to earn 68 points if "USA and/or Israel to execute an overt air strike against Iran" happens, instead of earning 32 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 32 points.
I choose to earn 71 points if "Gordon Brown to stop being Prime Minister, Labour leader or both" happens, instead of earning 29 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 29 points.
I choose to earn 73 points if "Firefox usage to reach 33% or higher" happens, instead of earning 27 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 27 points.
I choose to earn 75 points if "Named British Prince or Princess to marry publicly" happens, instead of earning 25 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 25 points.
I choose to earn 78 points if "No more than 2 major hurricanes in 2008 Atlantic Hurricane season" happens, instead of earning 22 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 22 points.
I choose to earn 79 points if "New England Patriots to win Super Bowl XLIII" happens, instead of earning 21 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 21 points.
I choose to earn 82 points if "John McCain to win at least 336 Electoral College seats" happens, instead of earning 18 points if it doesn't: players choosing to support the long-shot score zero, players choosing to oppose the long-shot score 18 points.
Accordingly a perfect score would have been available by only predicting the stock market drop long-shot and by refusing all the others. It would have scored 47+45+57+41+35+32+29+27+25+22+21+18=399. The highest score actually attained was 356, with the 43 points short of perfection due to incorrectly accepting the long-shots for hurricanes and the Patriots, and the winning player was quintus_marcius. Well done, David! David led almost from the start, though as he notes, he's played plenty of prediction games in the past, most frequently organised by gayparee, with no little success. This maiden victory is long overdue!
The minor placings are taken up by devjoe, scoring 297, missingdonut with 295, jaydlewis on 286, hedwig_snowy and his 266 then canadianpuzzler, at one point my own choice to win the whole thing, accruing 257 points. Scores were as low as 196, 193 and 162. I have a luridly coloured (but of course!) spreadsheet available with full scoring details.
I will note that I unofficially registered my predictions, which would have been good for 303 and thus sufficient for second place. However, the Rest Of The World who chose to oppose all twelve long-shots (for is making an entry selecting none of the choices different to not making an entry at all?) would have scored a moderately mighty 342. Possibly this does not reflect terribly well on all our performances, my own firmly included.
Thank you all for playing and apologies for the delayed results. I hope you enjoyed the game!
Reflecting on the game, I think my long-shots were slightly longer than I would have liked; they were intended to be about 20%-25% likely, thus making a result of only one turning out slightly disappointing, though probably not outside of tolerance. (If none of the twelve had happened, I would have been embarrassed.) Some long-shots were close - the Firefox one and, I still contend, the Microsoft-Yahoo! one being reasonably near misses - while others (oil prices, McCain seats) were spectacular misses. I rather enjoy that one of the long-shots was a miss for a reason very different to the ones we might have expected, but it's a shame it was the one regarding Half-Blood Prince after its postponement. It might have been interesting to run a two-tailed long-shot prediction game were you are able to predict long-shot low values as well as long-shot high ones, but I was deliberately trying to keep the complexity of the game low to attract players.
The concept of a long-shot prediction game still appeals to me, at least, simply because unlikely predictions are more interesting than easy-to-make ones. It would be very interesting to play a game that set out to reward those who could accurately distinguish a 10% long-shot from a 1% long-shot, but such a game would fairly necessarily be likely to be relatively complex - or, at least, rely on multiple trials of the long-shots in question, which is a facility that may not be available. It would be an attractive property to reward players' creativity in devising the long-shots, not least because they may well be more accurate at determining the length of a long-shot than I am. daweaver, as was, tried this in his 2006 prediction competition, which was closer to the mark, but I'm not sure it was quite there somehow. (Not least because I'm not sure what I'm looking for!)
Your thoughts and reflections would be welcome. I think I'll give this a rest for the remainder of the year and very probably next year as well, not least because I never really came up with a satisfactory name for the whole operation, though I may well participate in prediction games that any of you might run and would be delighted to learn of them. Besides, there are other - potentially more interesting - games I'd like to run online instead.
Current Mood: lazy
I don't have any useful thoughts or reflections, I'm afraid, except that I've been interested in following your exposition of the thinking behind it. And will be interested further if someone less lazy than me comes up with some contructive ideas.
Thank you for running the game, Chris. The aspect that I most liked was the multiple rounds. This format helped me to win by revealing how many people disagreed with me about the US topping the medals table. In more general terms, it created interesting tactical decisions regarding whether to vote for the events deemed especially unlikely. In a way it's sad that none of these happened, although had one of the ones that I had not picked done so I would no doubt once again be fuming about the luck element involved in these games.
Complete guesses do work!
Why oh why is Brown still PM!?!?!? :) I coulda been a mid-level contender! ;)
That was a really cool game, Chris! I'd love it if you ran another one!
Most interesting game, thanks for running it! Would Play Again!